Awesome
jsdiff
A JavaScript text differencing implementation. Try it out in the online demo.
Based on the algorithm proposed in "An O(ND) Difference Algorithm and its Variations" (Myers, 1986).
Installation
npm install diff --save
Usage
Broadly, jsdiff's diff functions all take an old text and a new text and perform three steps:
-
Split both texts into arrays of "tokens". What constitutes a token varies; in
diffChars
, each character is a token, while indiffLines
, each line is a token. -
Find the smallest set of single-token insertions and deletions needed to transform the first array of tokens into the second.
This step depends upon having some notion of a token from the old array being "equal" to one from the new array, and this notion of equality affects the results. Usually two tokens are equal if
===
considers them equal, but some of the diff functions use an alternative notion of equality or have options to configure it. For instance, by defaultdiffChars("Foo", "FOOD")
will require two deletions (o
,o
) and three insertions (O
,O
,D
), butdiffChars("Foo", "FOOD", {ignoreCase: true})
will require just one insertion (of aD
), sinceignoreCase
causeso
andO
to be considered equal. -
Return an array representing the transformation computed in the previous step as a series of change objects. The array is ordered from the start of the input to the end, and each change object represents inserting one or more tokens, deleting one or more tokens, or keeping one or more tokens.
API
-
Diff.diffChars(oldStr, newStr[, options])
- diffs two blocks of text, treating each character as a token.("Characters" here means Unicode code points - the elements you get when you loop over a string with a
for ... of ...
loop.)Returns a list of change objects.
Options
ignoreCase
: Iftrue
, the uppercase and lowercase forms of a character are considered equal. Defaults tofalse
.
-
Diff.diffWords(oldStr, newStr[, options])
- diffs two blocks of text, treating each word and each punctuation mark as a token. Whitespace is ignored when computing the diff (but preserved as far as possible in the final change objects).Returns a list of change objects.
Options
-
ignoreCase
: Same as indiffChars
. Defaults to false. -
intlSegmenter
: An optionalIntl.Segmenter
object (which must have agranularity
of'word'
) fordiffWords
to use to split the text into words.By default,
diffWords
does not use anIntl.Segmenter
, just some regexes for splitting text into words. This will tend to give worse results thanIntl.Segmenter
would, but ensures the results are consistent across environments;Intl.Segmenter
behaviour is only loosely specced and the implementations in browsers could in principle change dramatically in future. If you want to usediffWords
with anIntl.Segmenter
but ensure it behaves the same whatever environment you run it in, use anIntl.Segmenter
polyfill instead of the JavaScript engine's nativeIntl.Segmenter
implementation.Using an
Intl.Segmenter
should allow better word-level diffing of non-English text than the default behaviour. For instance,Intl.Segmenter
s can generally identify via built-in dictionaries which sequences of adjacent Chinese characters form words, allowing word-level diffing of Chinese. By specifying a language when instantiating the segmenter (e.g.new Intl.Segmenter('sv', {granularity: 'word'})
) you can also support language-specific rules, like treating Swedish's colon separated contractions (like k:a for kyrka) as single words; by default this would be seen as two words separated by a colon.
-
-
Diff.diffWordsWithSpace(oldStr, newStr[, options])
- diffs two blocks of text, treating each word, punctuation mark, newline, or run of (non-newline) whitespace as a token. -
Diff.diffLines(oldStr, newStr[, options])
- diffs two blocks of text, treating each line as a token.Options
ignoreWhitespace
:true
to ignore leading and trailing whitespace characters when checking if two lines are equal. Defaults tofalse
.ignoreNewlineAtEof
:true
to ignore a missing newline character at the end of the last line when comparing it to other lines. (By default, the line'b\n'
in text'a\nb\nc'
is not considered equal to the line'b'
in text'a\nb'
; this option makes them be considered equal.) Ignored ifignoreWhitespace
ornewlineIsToken
are also true.stripTrailingCr
:true
to remove all trailing CR (\r
) characters before performing the diff. Defaults tofalse
. This helps to get a useful diff when diffing UNIX text files against Windows text files.newlineIsToken
:true
to treat the newline character at the end of each line as its own token. This allows for changes to the newline structure to occur independently of the line content and to be treated as such. In general this is the more human friendly form ofdiffLines
; the default behavior with this option turned off is better suited for patches and other computer friendly output. Defaults tofalse
.
Note that while using
ignoreWhitespace
in combination withnewlineIsToken
is not an error, results may not be as expected. WithignoreWhitespace: true
andnewlineIsToken: false
, changing a completely empty line to contain some spaces is treated as a non-change, but withignoreWhitespace: true
andnewlineIsToken: true
, it is treated as an insertion. This is because the content of a completely blank line is not a token at all innewlineIsToken
mode.Returns a list of change objects.
-
Diff.diffSentences(oldStr, newStr[, options])
- diffs two blocks of text, treating each sentence as a token. The characters.
,!
, and?
, when followed by whitespace, are treated as marking the end of a sentence; nothing else is considered to mark a sentence end.(For more sophisticated detection of sentence breaks, including support for non-English punctuation, consider instead tokenizing with an
Intl.Segmenter
withgranularity: 'sentence'
and passing the result toDiff.diffArrays
.)Returns a list of change objects.
-
Diff.diffCss(oldStr, newStr[, options])
- diffs two blocks of text, comparing CSS tokens.Returns a list of change objects.
-
Diff.diffJson(oldObj, newObj[, options])
- diffs two JSON-serializable objects by first serializing them to prettily-formatted JSON and then treating each line of the JSON as a token. Object properties are ordered alphabetically in the serialized JSON, so the order of properties in the objects being compared doesn't affect the result.Returns a list of change objects.
Options
stringifyReplacer
: A custom replacer function. Operates similarly to thereplacer
parameter toJSON.stringify()
, but must be a function.undefinedReplacement
: A value to replaceundefined
with. Ignored if astringifyReplacer
is provided.
-
Diff.diffArrays(oldArr, newArr[, options])
- diffs two arrays of tokens, comparing each item for strict equality (===).Options
comparator
:function(left, right)
for custom equality checks
Returns a list of change objects.
-
Diff.createTwoFilesPatch(oldFileName, newFileName, oldStr, newStr[, oldHeader[, newHeader[, options]]])
- creates a unified diff patch by first computing a diff withdiffLines
and then serializing it to unified diff format.Parameters:
oldFileName
: String to be output in the filename section of the patch for the removalsnewFileName
: String to be output in the filename section of the patch for the additionsoldStr
: Original string valuenewStr
: New string valueoldHeader
: Optional additional information to include in the old file header. Default:undefined
.newHeader
: Optional additional information to include in the new file header. Default:undefined
.options
: An object with options.context
describes how many lines of context should be included. You can set this toNumber.MAX_SAFE_INTEGER
orInfinity
to include the entire file content in one hunk.ignoreWhitespace
: Same as indiffLines
. Defaults tofalse
.stripTrailingCr
: Same as indiffLines
. Defaults tofalse
.
-
Diff.createPatch(fileName, oldStr, newStr[, oldHeader[, newHeader[, options]]])
- creates a unified diff patch.Just like Diff.createTwoFilesPatch, but with oldFileName being equal to newFileName.
-
Diff.formatPatch(patch)
- creates a unified diff patch.patch
may be either a single structured patch object (as returned bystructuredPatch
) or an array of them (as returned byparsePatch
). -
Diff.structuredPatch(oldFileName, newFileName, oldStr, newStr[, oldHeader[, newHeader[, options]]])
- returns an object with an array of hunk objects.This method is similar to createTwoFilesPatch, but returns a data structure suitable for further processing. Parameters are the same as createTwoFilesPatch. The data structure returned may look like this:
{ oldFileName: 'oldfile', newFileName: 'newfile', oldHeader: 'header1', newHeader: 'header2', hunks: [{ oldStart: 1, oldLines: 3, newStart: 1, newLines: 3, lines: [' line2', ' line3', '-line4', '+line5', '\\ No newline at end of file'], }] }
-
Diff.applyPatch(source, patch[, options])
- attempts to apply a unified diff patch.Hunks are applied first to last.
applyPatch
first tries to apply the first hunk at the line number specified in the hunk header, and with all context lines matching exactly. If that fails, it tries scanning backwards and forwards, one line at a time, to find a place to apply the hunk where the context lines match exactly. If that still fails, andfuzzFactor
is greater than zero, it increments the maximum number of mismatches (missing, extra, or changed context lines) that there can be between the hunk context and a region where we are trying to apply the patch such that the hunk will still be considered to match. Regardless offuzzFactor
, lines to be deleted in the hunk must be present for a hunk to match, and the context lines immediately before and after an insertion must match exactly.Once a hunk is successfully fitted, the process begins again with the next hunk. Regardless of
fuzzFactor
, later hunks must be applied later in the file than earlier hunks.If a hunk cannot be successfully fitted anywhere with fewer than
fuzzFactor
mismatches,applyPatch
fails and returnsfalse
.If a hunk is successfully fitted but not at the line number specified by the hunk header, all subsequent hunks have their target line number adjusted accordingly. (e.g. if the first hunk is applied 10 lines below where the hunk header said it should fit,
applyPatch
will start looking for somewhere to apply the second hunk 10 lines below where its hunk header says it goes.)If the patch was applied successfully, returns a string containing the patched text. If the patch could not be applied (because some hunks in the patch couldn't be fitted to the text in
source
),applyPatch
returns false.patch
may be a string diff or the output from theparsePatch
orstructuredPatch
methods.The optional
options
object may have the following keys:fuzzFactor
: Maximum Levenshtein distance (in lines deleted, added, or subtituted) between the context shown in a patch hunk and the lines found in the file. Defaults to 0.autoConvertLineEndings
: Iftrue
, and if the file to be patched consistently uses different line endings to the patch (i.e. either the file always uses Unix line endings while the patch uses Windows ones, or vice versa), thenapplyPatch
will behave as if the line endings in the patch were the same as those in the source file. (Iffalse
, the patch will usually fail to apply in such circumstances since lines deleted in the patch won't be considered to match those in the source file.) Defaults totrue
.compareLine(lineNumber, line, operation, patchContent)
: Callback used to compare to given lines to determine if they should be considered equal when patching. Defaults to strict equality but may be overridden to provide fuzzier comparison. Should return false if the lines should be rejected.
-
Diff.applyPatches(patch, options)
- applies one or more patches.patch
may be either an array of structured patch objects, or a string representing a patch in unified diff format (which may patch one or more files).This method will iterate over the contents of the patch and apply to data provided through callbacks. The general flow for each patch index is:
options.loadFile(index, callback)
is called. The caller should then load the contents of the file and then pass that to thecallback(err, data)
callback. Passing anerr
will terminate further patch execution.options.patched(index, content, callback)
is called once the patch has been applied.content
will be the return value fromapplyPatch
. When it's ready, the caller should callcallback(err)
callback. Passing anerr
will terminate further patch execution.
Once all patches have been applied or an error occurs, the
options.complete(err)
callback is made. -
Diff.parsePatch(diffStr)
- Parses a patch into structured dataReturn a JSON object representation of the a patch, suitable for use with the
applyPatch
method. This parses to the same structure returned byDiff.structuredPatch
. -
Diff.reversePatch(patch)
- Returns a new structured patch which when applied will undo the originalpatch
.patch
may be either a single structured patch object (as returned bystructuredPatch
) or an array of them (as returned byparsePatch
). -
Diff.convertChangesToXML(changes)
- converts a list of change objects to a serialized XML format -
Diff.convertChangesToDMP(changes)
- converts a list of change objects to the format returned by Google's diff-match-patch library
Universal options
Certain options can be provided in the options
object of any method that calculates a diff (including diffChars
, diffLines
etc. as well as structuredPatch
, createPatch
, and createTwoFilesPatch
):
-
callback
: if provided, the diff will be computed in async mode to avoid blocking the event loop while the diff is calculated. The value of thecallback
option should be a function and will be passed the computed diff or patch as its first argument.(Note that if the ONLY option you want to provide is a callback, you can pass the callback function directly as the
options
parameter instead of passing an object with acallback
property.) -
maxEditLength
: a number specifying the maximum edit distance to consider between the old and new texts. You can use this to limit the computational cost of diffing large, very different texts by giving up early if the cost will be huge. This option can be passed either to diffing functions (diffLines
,diffChars
, etc) or to patch-creation function (structuredPatch
,createPatch
, etc), all of which will indicate that the max edit length was reached by returningundefined
instead of whatever they'd normally return. -
timeout
: a number of milliseconds after which the diffing algorithm will abort and returnundefined
. Supported by the same functions asmaxEditLength
. -
oneChangePerToken
: iftrue
, the array of change objects returned will contain one change object per token (e.g. one per line if callingdiffLines
), instead of runs of consecutive tokens that are all added / all removed / all conserved being combined into a single change object.
Defining custom diffing behaviors
If you need behavior a little different to what any of the text diffing functions above offer, you can roll your own by customizing both the tokenization behavior used and the notion of equality used to determine if two tokens are equal.
The simplest way to customize tokenization behavior is to simply tokenize the texts you want to diff yourself, with your own code, then pass the arrays of tokens to diffArrays
. For instance, if you wanted a semantically-aware diff of some code, you could try tokenizing it using a parser specific to the programming language the code is in, then passing the arrays of tokens to diffArrays
.
To customize the notion of token equality used, use the comparator
option to diffArrays
.
For even more customisation of the diffing behavior, you can create a new Diff.Diff()
object, overwrite its castInput
, tokenize
, removeEmpty
, equals
, and join
properties with your own functions, then call its diff(oldString, newString[, options])
method. The methods you can overwrite are used as follows:
castInput(value, options)
: used to transform theoldString
andnewString
before any other steps in the diffing algorithm happen. For instance,diffJson
usescastInput
to serialize the objects being diffed to JSON. Defaults to a no-op.tokenize(value, options)
: used to convert each ofoldString
andnewString
(after they've gone throughcastInput
) to an array of tokens. Defaults to returningvalue.split('')
(returning an array of individual characters).removeEmpty(array)
: called on the arrays of tokens returned bytokenize
and can be used to modify them. Defaults to stripping out falsey tokens, such as empty strings.diffArrays
overrides this to simply return thearray
, which means that falsey values like empty strings can be handled like any other token bydiffArrays
.equals(left, right, options)
: called to determine if two tokens (one from the old string, one from the new string) should be considered equal. Defaults to comparing them with===
.join(tokens)
: gets called with an array of consecutive tokens that have either all been added, all been removed, or are all common. Needs to join them into a single value that can be used as thevalue
property of the change object for these tokens. Defaults to simply returningtokens.join('')
.postProcess(changeObjects)
: gets called at the end of the algorithm with the change objects produced, and can do final cleanups on them. Defaults to simply returningchangeObjects
unchanged.
Change Objects
Many of the methods above return change objects. These objects consist of the following fields:
value
: The concatenated content of all the tokens represented by this change object - i.e. generally the text that is either added, deleted, or common, as a single string. In cases where tokens are considered common but are non-identical (e.g. because an option likeignoreCase
or a customcomparator
was used), the value from the new string will be provided here.added
: true if the value was inserted into the new string, otherwise falseremoved
: true if the value was removed from the old string, otherwise falsecount
: How many tokens (e.g. chars fordiffChars
, lines fordiffLines
) the value in the change object consists of
(Change objects where added
and removed
are both false represent content that is common to the old and new strings.)
Examples
Basic example in Node
require('colors');
const Diff = require('diff');
const one = 'beep boop';
const other = 'beep boob blah';
const diff = Diff.diffChars(one, other);
diff.forEach((part) => {
// green for additions, red for deletions
let text = part.added ? part.value.bgGreen :
part.removed ? part.value.bgRed :
part.value;
process.stderr.write(text);
});
console.log();
Running the above program should yield
<img src="images/node_example.png" alt="Node Example">Basic example in a web page
<pre id="display"></pre>
<script src="diff.js"></script>
<script>
const one = 'beep boop',
other = 'beep boob blah',
color = '';
let span = null;
const diff = Diff.diffChars(one, other),
display = document.getElementById('display'),
fragment = document.createDocumentFragment();
diff.forEach((part) => {
// green for additions, red for deletions
// grey for common parts
const color = part.added ? 'green' :
part.removed ? 'red' : 'grey';
span = document.createElement('span');
span.style.color = color;
span.appendChild(document
.createTextNode(part.value));
fragment.appendChild(span);
});
display.appendChild(fragment);
</script>
Open the above .html file in a browser and you should see
<img src="images/web_example.png" alt="Node Example">Example of generating a patch from Node
The code below is roughly equivalent to the Unix command diff -u file1.txt file2.txt > mydiff.patch
:
const Diff = require('diff');
const file1Contents = fs.readFileSync("file1.txt").toString();
const file2Contents = fs.readFileSync("file2.txt").toString();
const patch = Diff.createTwoFilesPatch("file1.txt", "file2.txt", file1Contents, file2Contents);
fs.writeFileSync("mydiff.patch", patch);
Examples of parsing and applying a patch from Node
Applying a patch to a specified file
The code below is roughly equivalent to the Unix command patch file1.txt mydiff.patch
:
const Diff = require('diff');
const file1Contents = fs.readFileSync("file1.txt").toString();
const patch = fs.readFileSync("mydiff.patch").toString();
const patchedFile = Diff.applyPatch(file1Contents, patch);
fs.writeFileSync("file1.txt", patchedFile);
Applying a multi-file patch to the files specified by the patch file itself
The code below is roughly equivalent to the Unix command patch < mydiff.patch
:
const Diff = require('diff');
const patch = fs.readFileSync("mydiff.patch").toString();
Diff.applyPatches(patch, {
loadFile: (patch, callback) => {
let fileContents;
try {
fileContents = fs.readFileSync(patch.oldFileName).toString();
} catch (e) {
callback(`No such file: ${patch.oldFileName}`);
return;
}
callback(undefined, fileContents);
},
patched: (patch, patchedContent, callback) => {
if (patchedContent === false) {
callback(`Failed to apply patch to ${patch.oldFileName}`)
return;
}
fs.writeFileSync(patch.oldFileName, patchedContent);
callback();
},
complete: (err) => {
if (err) {
console.log("Failed with error:", err);
}
}
});
Compatibility
jsdiff supports all ES3 environments with some known issues on IE8 and below. Under these browsers some diff algorithms such as word diff and others may fail due to lack of support for capturing groups in the split
operation.
License
See LICENSE.
Deviations from the published Myers diff algorithm
jsdiff deviates from the published algorithm in a couple of ways that don't affect results but do affect performance:
- jsdiff keeps track of the diff for each diagonal using a linked list of change objects for each diagonal, rather than the historical array of furthest-reaching D-paths on each diagonal contemplated on page 8 of Myers's paper.
- jsdiff skips considering diagonals where the furthest-reaching D-path would go off the edge of the edit graph. This dramatically reduces the time cost (from quadratic to linear) in cases where the new text just appends or truncates content at the end of the old text.